Community Size-Structure of the Northwest Atlantic Groundfish Communities, Response to Direct Disturbance and a Changing Environment

Size Spectra Analysis of the Northeast US Finfish Community

Author
Affiliation

Gulf of Maine Research Institute

Published

August 15, 2023

Potential Journals:

  1. ICES Journal of Marine Science

Abstract

The surface waters of the Northwest Atlantic Ocean are among the fastest warming on Earth. This area is highly-productive biologically, and there are concerns that ecological consequences will follow this rapid-warming. Research on the impacts of this rapid warming has primarily focused on high-profile and/or upper trophic level species. Ecological theory and laboratory studies suggest that elevated temperatures facilitate early maturation and smaller adult body-sizes. However, it is unclear whether that relationship might be mitigated against through adaptive behaviors in an open ocean environment. Here we’ve investigated ecosystem wide impacts on the individual size distribution (ISD) to track changes in community size structure. In cases where community responses are not adequate to counter the impacts of elevated temperatures, we anticipated a steepening of the size spectrum slope (ISD exponent). A steeper relationship relating to a reduction in larger sized individuals and an increased prevalence of smaller sized individuals. Using data from fisheries independent surveys we calculated the community size spectra for four regions along the US NE continental shelf. Correlation/regression analyses were then performed to then assess the degree to which these changes were in alignment to hypothesized bottom-up and top-down disturbances. At the regional scale, we found that community size structure changes (spectra slope) were the largest in the Northern regions, in the Gulf of Maine and Georges Bank. These areas are home to the coldest temperatures and the largest proportions of groundfish species in the community. Spectrum slope declines were most pronounced in the 80’s and 90’s, before the rapid warming of the last decade. The timing of these declines suggest that external factors drove the initial declines of larger-sized individuals within the communities, before elevated temperatures began to influence the ecosystem. Correlation analyses reveal that while fisheries landings are strongly correlated with these declines, bottom-up factors of zooplankton community metrics, Gulf Stream Index, and SST anomalies are also important. While the primary pressure of fisheries exploitation has declined dramatically over time, the recovery of larger-sized individuals has not been seen. That kind of recovery will likely depend on the elevated temperatures seen over the last decade.

Introduction

Understanding the impacts of climate change on local and regional scales is a high priority area of study among the international scientific community. The largest determinant of annual climatic timescale variation is the heat exchange between the atmosphere and the ocean Glantz et al. (1991). Major atmospheric and oceanographic climate modes like the El Nino Southern Oscillation (ENSO), the Gulf Stream Index (GSI) or the North Atlantic Oscillation (NAO) can be linked to localized climate impacts via energy transport between the ocean and atmosphere in a linkage referred to as a “teleconnection” Liu and Alexander (2007). The existence of these pathways demands that scientists understand the inter-connected relationship between local weather and large-scale oceanographic regimes. Scientists following climate change through study of Earth’s energy imbalance have observed that over 90% of that excess heat from climate change is accumulating in the ocean Meyssignac et al. (2019). This fact has re-focused attention on concerns for the future health of marine ecosystems as temperatures, acidification, sea level, and ocean circulation patterns change (“Warming trends increasingly dominate global ocean | nature climate change” (n.d.); Findlay and Turley (2021); University of South Carolina et al. (2021); Frederikse et al. (2020); Neto et al. (2021)). Focusing more narrowly on the impacts of rapid temperature change, scientists have predicted that ecological communities will respond in a number of predictable pathways, each to the limit that their constituent species’ life histories can support a response. These responses include shifts in geographic distribution annd alterations to their growth and maturation (Bergmanns Rule, Blackburn et al. (1999), CITE early maturation when warm, Atkinson (1995) ) . Marine ecosystems are also shaped directly through human interactions like fishing, which has also been shown to impact these single-species characteristics ( CITE: genetic impacts of size-selective gear, allee effects).

Marine ecosystems are complex and are difficult to sample accurately. Much research on the early impacts of climate change on marine systems has focused on individual species or small communities that can be consistently sampled and/or have their interactions explicitly described ( Pinsky et al. (2013), Nye et al. (2009), Pershing et al. (2015) ). An obvious weakness of these approaches is the assumption that all ecological needs provided through inter-species interactions will be equally met under projected climate change conditions. Another lens with which to study the organization of a community is through the study of traits. Size is commonly referenced as the master trait, with well documented relationships with metabolism and bio-energetics Sheldon et al. (1972) Brown et al. (2000) White et al. (2007) . Marine communities are very-often size-structured, with size correlating to trophic position, making them ideal for study through this lens. Size spectra theory is an ecological framework for studying the relationship between the abundance and size of individuals in a community . This relationship relates how efficiently energy flows from the smallest individuals at the lowest trophic levels to larger individuals occupying higher trophic levels Andersen and Beyer (2006) Hillaert et al. (2018) . It is a taxon-agnostic method for detecting changes to ecosystem health and has shown to be sensitive to both large-scale environmental disturbance as well as direct human interaction Shin et al. (2005) Kerr and Dickie (2001) . When organized on a log scale, this relationship between abundance and size reveals a linear decline in abundance with increasing body size White et al. (2008) . Highly productive and diverse marine ecosystems are associated with less steep declines in abundance with size, and support larger numbers of higher trophic level individuals. By studying how this relationship changes through time scientists can observe the energy-transfer efficiency of a community as a signal of ecological health and robustness to external changes in the environment Sprules and Barth (2016) .

Our study is focused on changes in the size distribution finfish community of the Northeast US continental shelf. This is a region that is experiencing rapid increases to temperature tied to changes in the behaviors of the Gulf Stream and Labrador Currents University of South Carolina et al. (2021) . This ecosystem has been sampled extensively as part of national fisheries management efforts and has long-term records on the size distribution of the finish community, providing a long-history with which to study changes in size structure. Based on early observational studies and ecological theory we anticipate that individual level responses to a rapidly warming environment will be detectable in the size-structure of the region’s ecological community, evidence that ecosystems are experiencing change-of-function as a result of rapid environmental change.

Rapid Warming of Northeast Shelf

The Northwest Atlantic is one of the fastest warming locations in the global oceans. Sea surface temperatures in the Gulf of Maine since 1982 have been warming at rates faster than 96% of the world’s oceans, with similar warming rates along the northwest Atlantic shelf (Pershing et al. (2018)). The persistent elevated temperature regime of the area is a result of several forces: a combination of shifting ocean currents and the unique bathymetry of the region. A Northward shift in the Gulf Stream directly increased the regional temperatures through increased transport of warm Gulf Stream water into areas like the Gulf of Maine. The Northward Gulf Stream shift is associated with a higher frequency of warm core rings, and the obstruction of cold-water Scotian Shelf current flow that otherwise counters the influence of the Gulf Stream on the region’s temperatures (Gangopadhyay et al. (2019); Meyer-Gutbrod et al. (2021)). The combination of these oceanographic changes has led to a warmer continental shelf habitat, which has caused alarm among scientists studying the ecology of the region.

Temperature plays a critical role on biological function, affecting many of the chemical reactions that underpin basic physiological function. Reactions impacting critical actions like locomotion and feeding behavior, metabolism and development, and even maturation rates. Because of these relationships, species have evolved thermal preferences around which these functions each operate efficiently. Individuals that are unable to maintain their thermal preferences internally risk metabolic costs unless they can follow their thermal preference in the environment through locomotion or adapt to less-favorable conditions through changes in behavior. In an era of anthropogenic climate change, there is an expectation that many species will be displaced from historic habitats in their efforts to follow their thermal preferences. Recent research in marine environments has shown evidence of this, as species are now shifting to higher latitudes and to deeper depths in the pursuit of more favorable conditions (Nye et al. (2009); Pinsky et al. (2013)). Other work suggests that temperature related impacts may manifest through physiological changes and changes in seasonal phenology, hindering species conservation efforts (Miller et al. (2018); Pershing et al. (2015); Meyer-Gutbrod et al. (2021)). Need to connect temperature to size here: Direct quote from Guiet et al. (2016) , but nails the connection back to temp expectations:

Because it controls chemical reactions, temperature controls metabolic rates which underpin maintenance, growth or reproduction (Clarke and Johnston, 1999; Kooijman, 2010) as well as the functional responses to food density (Rall et al., 2012). Guiet et al. (2016)… In addition to the impact of temperature on communities’ intercepts (heights), the impact of temperature on the speed of the energy flow within communities may affect other properties, such as their resilience to perturbations or the intensity of trophic cascades (Andersen and Pedersen, 2009).

The potential for elevated temperatures to impact the size structure of an ecosystem has implications for the ecosystem resilience in the face of climate change, as well as the blue economies & natural resource systems that rely upon their good health.

Size Spectra Theory

Size is a defining characteristic of species and mediates many ecological interactions and metabolic pathways (Brown et al. 2000) . Size is a major determining factor for the mobility of an organism which directly impacts the ability to evade predation, find foraging success, and efforts to locate and follow seasonal habitats. Size and physical dimensions also impact the metabolic costs associated with each of these behaviors (Hillaert et al. 2018), mediating exchange rates with the immediate environment like heat loss or desiccation in terrestrial species (Gillooly et al. (2001); Heatwole et al. 1969). Body size even informs life history features like life span and the trophic position an individual might occupy through its impact on metabolism and resource use (White et al. 2007). Size structured environments are a fundamental organizational pattern that emerges from these relationships add_citation. Within strongly size-structured ecosystems, growth and maturity changes alter fitness and ultimately determine whether a species is successful in that environment add_citation . Ecological theory is rich with models relating how energy transfers from smaller prey species to larger predatory trophic levels, the allocation of energy for growth, and the trade offs of allocating energy towards those ends (Bertalanffy (1938); Bertalanffy (1957)); add_size_theory_citations). A globally persistent pattern in ecology entangled in those relationships and their critiques is the decline in abundance with increasing body size (Damuth (1981); Currie (1993); Sheldon et al. (1972); Loeuille and Loreau (2006)). The relationship between size and abundance integrates processes operating on the cellular, individual, and community levels simultaneously. By measuring how this relationship changes, scientists can detect whether the fundamental energy transfer pathways from small individuals to larger predators have been impacted or altered, an outward indication of ecosystem-wide changes. The quantities for size and abundance are often the most readily collected data assets of any ecological community. This creates an opportunity to learn much about a system from a relatively low-effort in data collection. For these reasons, size spectrum analyses and individual size distribution (ISD) methods have gained increasing attention as an entry point to assessing ecosystem health and to detect system-wide disturbance (Shin et al. (2005)‘; Pomeranz et al. (2022); White et al. (2007)). Spectra models are also taxon agnostic and avoid the need to explicitly articulate each predator-prey interaction. The “size spectrum” describes the distribution of abundance or biomass as a function of individuals’ mass on a log–log scale (Guiet et al. (2016); Kerr and Dickie (2001)) . Size spectra condense the complexities of predator prey networks and their interactions into taxon-agnostic size-based indices. These indices capture the emergent properties of a system, and have become increasingly used as indicators of ecosystem health. Within the context of fisheries management, changes in spectrum slopes have been associated with fishing exploitation, primarily through the targeted removal of larger individuals (Bianchi et al. (2000); Shin et al. (2005)). Numerical experiments have also linked changes in slope to environmental disturbances (Guiet et al. (2016)). Size spectra have also been shown to express predictable relationships between ecosystems of similar productivity levels as well as from distinct temperature regimes (Guiet et al. (2016)). Use Sprules and Barth Paper to discuss applications Use Pomerantz paper & Edwards to extend into ISD

External Drivers of Size Structure

Our study area’s ecology is one that has historically been studied through the fishing grounds it once supported and their challenges. The focus has traditionally been on the success/failure of single species management actions, however there are growing efforts to better understand the full ecological impact of this endeavor. Fishing practices are often size-selective, with harvesters targeting larger individuals for a higher yield on any time and capital invested. Larger individuals are also commonly the older individuals in a population, having lived longer to achieve those sizes. With the unfortunate side-effect of having an outsized impact on the reproductive potential of the population. Larger individuals have a greater impact on population resilience and recovery, capable of holding more (and often of higher quality) eggs which increase the odds of recruitment success. Heavily fished populations have been shown to exhibit changes to growth rates, suggesting an additional genetic impact from size-selective fishing on the target-species. The direct removal of larger individuals carries a predictable impact on the size distributions of the target species, but may also change that of non-target species as well. Non-target species may be impacted as by-catch from the fishery, or may alternatively find relief from predation if the target species is a direct predator. This can even create circumstances where prey species can crowd out, or outcompete the juveniles of their natural predators, reinforcing the new community. Industrial fishing practices are inherently size-selective, with an outsized footprint relative to small-scale fisheries. Large removals have an immediate and measurable impact on the community size-distribution with potential additional impacts on the future population as well. Size-based harvest in fisheries has been shown to create selective pressures that promote characteristics of early maturation at smaller sizes add_citation. These consequences may independently or in combination create circumstances where a community is unable to return to conditions before these disturbances occurred. reference gb/ss spectra early work (Duplisea & Kerr 1995, or Kerr book)

The ecological community in our study area has with no uncertainty been altered through human behaviors over the last century. Early research estimated that biomass had more than halved in some areas by the 1960’s, pre-dating federal monitoring efforts and inspiring their formation (Fogarty and Murawski (1998)). Key stocks that supported international fishing efforts collapsed, and recovery efforts have in many species fallen short. Work exploring retrospective patterns in stock assessment predictions have highlighted that these shortfalls can be partially attributed to un-accounted for influences of a changing environment. However, other research has noted that the community structure itself may have changed, and with it, how it responds to outside pressure. That same early research noted that species replacement of commercial target species by skate and dogfish was happening as their numbers quickly began to rise. A wealth of anecdotal evidence and records from commercial landings support the conclusion that the region does not support either the fisheries or the fishing communities in the same way today. The Lobster industry remains King in the Gulf of Maine, another side-effect of the decline in large predators in the area. With other similar declines in large-fish populations up and down the coast, scientists have recently brought the focus of their attention away from single-species management to more holistic and ecosystem-wide approaches.

Regime Shifts in Marine Systems

Goals:

Want to hit: dynamic tipping points, hysteresis, and non-stationary functional relationships. Show examples of the idea on single species, suggest that we size spectra can be used to check the whole community for tipping points

When an ecosystem experiences an abrupt change in its internal dynamics in response to internal or external pressures, it is said to have crossed a tipping point. When these tipping points are crossed, it often becomes difficult or impossible to return to the previous state. This is true even when the pressures that caused the shift have ceded and favorable conditions have returned, a phenomenon known as hysteresis. This phenomenon has been observed in individual population trends, and is a major factor undermining sustainable management of fisheries (Blocker 2023; Sguottti 2019) . Another feature that may accompany a regime shift is the altering of functional relationships with other biotic or abiotic features of the environment. When this occurs, the productivity of a species and/or its resilience to stress may be altered, and assumptions underpinning sustainable harvest thresholds should be updated (Blocker 2023). At the community level the alternative community states may support a different composition of species. The community itself may respond differently to the same stressors/stimulus. In the Northeast Shelf Region there is evidence to suggest that through a combination of human influences and environmental changes that the ecology of some areas may have crossed ecological tipping points. This study explores whether evidence for an ecological tipping point can be seen through the perspective of the size spectrum, and whether there is evidence for hysteresis, or non-stationarity in the community’s relationships to external forces.

Purpose

Understanding that human populations depend on the health of their ecosystems, there is a need to better understand when we’ve crossed ecological tipping points. One way to capture these changes is through the use of community-wide metrics that capture the inner-efficiencies of many trophic interactions. Size based indices are metrics that can be estimated from the information that has historically been available from long-term survey efforts. These indices have been shown to be sensitive to the impacts of fishing, but also capture environmental dynamics as well. We estimated size spectrum relationships to use as size based indicators of the ecological communities within each sub-region of the Northeast US continental shelf. Based on ecological theory we believe that the sustained increases in temperature in the NW Atlantics should have a physiological impact on the community size structure. We hypothesize that rapid warming alters the community through the direct influence of temperature on metabolism, growth, and population productivity. We also hypothesize that size spectra will uniquely reflect the physical environment of the regions, with less steep size spectra in regions with colder temperatures and of higher primary productivity. Understanding that temperature is not the sole mechanism for altering community size structure, we will explore how the temporal structures in size spectra have changed in relation to major external drivers. We posit that distinct patterns of deviation from that signature may reflect a fundamental change in the ecology of that region, an indication of a possible ecological regime change, and an altered relationship to human and environmental disturbances.

Methods

Fish Data Source and Processing

Data on the biomass, abundance, and size of fish on the Northeast U.S. Shelf were collected as part of the Northeast Fisheries Science Center’s bottom trawl survey (Grosslein 1969, Azarovits 1981, Politis 2014). This survey is conducted from Cape Hatteras, North Carolina to the Gulf of Maine each year in the spring and in the fall. The survey follows a stratified random sampling design, with strata defined based on depth, bottom habitat, and latitude. Trawls are performed for a fixed duration at each station, reporting aggregate abundance and biomass for all species caught, and measuring individual lengths and weights for the catch of each species or a sub-sample if that catch is large. Correction factors were applied to aggregate species abundance and biomass to account for changes in vessels, gear, and doors used in the survey over time (Sissenwine and Bowman 1978, Byrne and Forrester 1991, Miller et al. 2010). However, abundance and biomass at length needed to be estimated after these aggregate corrections. As such, abundance at length for each species was adjusted to match the corrected aggregate species abundance at each station, such that for each species, the sum of the resulting estimated abundance numbers across each length is equal to the corrected aggregate abundance.

Community Composition & Functional Groups

Analyses were performed using 68 species. These species were selected based on the availability of published weight-at-length relationships (Wigley et al. 2003) and represented 98.98% of the total biomass caught in the survey. Each species was assigned to a functional group based on life history and geography using the definitions of (Hare et al. (2016)). Functional groups included coastal fishes, diadromous fishes, elasmobranchs, groundfish, pelagic fishes, and reef fish species (Table 1.). Six species with available length-weight details did not have a functional group designation, these species were designated as reef species. Exploratory analyses showed that the pelagic species biomass was low in all regions, and is unlikely to be representative of true biomass trends due to gear selectivity.

Published length-weight relationships ( Wigley et al. 2003 ) were used to convert from length data, available for all individuals, into their corresponding biomass-at-length. To account for differences in sampling effort among survey strata, all corrected abundance-at-length data were area-stratified. Area-stratified biomass-at-length values were then computed as the product of area-stratified abundance-at-length and estimated weight-at-length. All analyses were performed using area-stratified abundances and their associated area-stratified biomass estimates, hereby referred to as simply abundance-at-length & biomass-at-length

Community Metrics

Our analyses used all data collected during the spring and fall surveys from 1970-2019. Data were grouped using survey-design strata into four sub-regions: Gulf of Maine, Georges Bank, Southern New England, Mid-Atlantic Bight (Figure 1.). These sub-regions have been widely used in regional ecological studies (e.g., ). For each region, we developed the following time series of ecological indicators:

  1. Annual mean abundance and biomass by functional group

  2. Annual mean length and weight of the aggregate community and for each functional group

  3. Annual estimates of the community size spectrum slope

Quantifying Body Size Changes

Changes in the size structure of the community was measured using the average length and weight across all species, each weighted by their species and size-specific catch rates. The average body length (cm) and body weight (kg) was calculated for each region and within each functional group.

\[X_j = \frac{ \sum_{}^{} n_iX_{ij} }{ \sum~w_i }\] Where \(X\) is the body-size metric, \(j\) is the year, & \(n\) is the abundance for each station \(i\).

Data for body size trends were not truncated using any minimum or maximum size and reflect all available catch data for the 68 species in this study for which biomass-at-length could be estimated.

Estimating Size Spectra

Abundance size spectra were calculated using the area-weighted abundance at length information from the catch data. Lengths of individuals in the catch data are measured to the nearest cm, with smaller specimens measured to the nearest millimeter. Because individual biomass is estimated from those length measurements, there is a range of possible body mass values between the increments used (cm and mm). The relationship between length and mass in fishes is exponential and taxon specific. Using only the lower or upper end of those ranges introduces biases which are different for each species and increase with larger body-sizes. To account for these biases, we used the extended likelihood method (MLEbin) of Edwards et al. (2020). This method estimates the exponent of size spectra (b) for a bounded power law relationship between abundance and length-estimated biomass. The exponent of this relationship is analogous to slope estimates from linear regression on logarithmic axes, with a steeper slope (i.e. a more negative b) indicating fewer large-bodied individuals and/or more small-bodied fishes (Carvalho 2021) . This method has been shown to be the most accurate for estimating the exponent of size spectra when tested against alternative methods that require arbitrary decisions around binning the data and make estimates difficult to compare directly across studies (White 2008; Sprules & Barth 2016; Edwards 2017).

\[ \begin{align*} f(x) = \frac{ (\lambda + 1)x^{\lambda} }{ x^{\lambda+1}_{max} - x^{\lambda+1}_{min} }~~~~~~\lambda\neq1, \\ \\ f(x) = \frac{1}{logx_{max} - logx_{min}}~~~~~~\lambda=1 \end{align*} \tag{1}\]

Using this method, the size spectra exponent (b) was estimated for each region from 1970 to 2019. A minimum biomass of 1g was used for the lower bound and a maximum biomass of 10kg was used as an upper bound for the ISD’s bounded power law probability density function Equation 1, where X is body mass & Λ is the scaling exponent of the abundance density function. This body-size range constitutes 97.83% of all the biomass for the 68 species included in this study. This body-size range truncation was done to account for poor gear selectivity at the smallest and largest size ranges. This measure also imposes shared bounds to the size range covered by our size spectra, reflecting only the ranges we’d expect to consistently sample across these different areas. Exponents of size spectra (b) were calculated using code modified from the sizeSpectra package (Edwards et al. (2017); Edwards et al. (2020)) and implemented using the R statistical programming language.

Drivers of Size Spectra Changes

The impact of external factors on the changes in size spectra were explored using multiple regression analyses. Annual variation in size spectrum slopes was modeled using several hypothesized drivers including the environmental state and measures of anthropogenic disturbances. The large-scale environmental drivers included were regional SST anomalies & the larger-scale impact of the Gulf Stream Index (GSI). Fishing pressure represents the primary top-down anthropogenic driver in the region, and was investigated using the aggregate regional commercial landings data. Bottom-up ecological interactions were explored using zooplankton indices from the EcoMon plankton survey. Each of these drivers (climate, productivity, fishing) have independently been shown in other works to have measurable impacts on size spectra. Our exploratory investigations here are an effort to evaluate which best explain the trends within our study region, and whether those relationships have changed over the study period.

Gulf Stream Index

Data for the Gulf stream index (GSI) was obtained from the ecodata package in R (Bastille & Hardison 2018) . This package supplies GSI data at monthly intervals following the methodology of Pérez-Hernández and Joyce (2014) and Joyce et al. (2019) , using as sea level height anomaly data from the Copernicus Marine Environment Monitoring Service.

Sea Surface Temperature Data

Global sea surface temperature (SST) data were obtained via NOAA’s Optimum Interpolation SST analysis (OISSTv2), which provides daily SST values at a 0.25° latitude x 0.25° longitude resolution (Reynolds et al. 2007). A daily climatology for every 0.25° pixel in the global data set was created using average daily temperatures spanning the period of 1982-2011. Daily anomalies were then computed as the difference between observed temperatures and the daily climatological average. OISSTv2 data used in these analyses were obtained from the NOAA Physical Sciences Laboratory, Boulder, Colorado, USA from their website at https://psl.noaa.gov/data/gridded/data.noaa.oisst.v2.html.

SST data were regionally averaged to match the four survey sub-regions described above (Figure 1a), producing daily time series for each area. These were then averaged into annual time series of surface temperatures and anomalies. All averaging between satellite grid units was done with area-weighting of the grid cells to account for differences in cell size with latitude/longitude in the OISSTv2 data.

Commercial Fishing

Fishing pressure in the region was indexed using state and federal commercial fishing landings. These data were obtained from the Greater Atlantic Regional Fisheries Office (GARFO) for statistical areas that are routinely used for fisheries reporting and management (Figure 1., right). Individual statistical areas were aggregated into regions that closely align with the survey areas we defined for the size spectra analyses (Figure 1 b.). Each region’s commercial landings timeline was scaled using z-score normalization

Zooplankton Indices

Zooplankton data was from the National Oceanographic and Atmospheric Administration Marine Resources Monitoring, Assessment and Prediction (MARMAP) program and Ecosystem Monitoring (EcoMon) cruises detailed extensively in Kane (2007), Kane (2011), and Morse et al. (2017). Abundance anomalies were computed from the expected abundance on the day of sample collection. The small copepod index is computed by averaging the individual abundance anomalies of Pseudocalanus spp., Centropages hamatus, Centropages typicus, and Temora longicornis. The “large-copepod” anomaly values are the abundance anomaly of Calanus finmarchicus, the largest copepod in the Northeast U.S. region. Anomalies for both large and small zooplankton groups were averaged within ecological production units by the data provider, producing annual timeseries for the Gulf of Maine, Georges Bank, and Mid-Atlantic Bight. The timeseries for the Mid-Atlantic Bight EPU was used as a predictor for both the Mid-Atlantic Bight and Southern New England regions in our analyses. This decision was made to address how the spatial coverage of this EPU spans the joint area of our two southernmost regions.

Exploring Influence of Drivers on Size Spectra

The hypothesized drivers for size spectrum changes were evaluated using multiple regression analysis. Candidate models for each region were informed by the any observed temporal structures in the size spectra changes through time, and with the inclusion of important lag-effects of predictors.These modeling decisions were informed by the following steps performed prior to multiple regression analysis with the above drivers.

Temporal Lags in Hypothesized Drivers

In addition to exploring their contemporaneous impacts, the importance of any time-lags on predictors was also evaluated. Important lagged relationships were identified through the use of cross correlation function (CCF) estimates. These estimates looked at correlations at one-year lag intervals to investigate potential time-delayed effects between the hypothesized drivers and the community’s size spectra. CCF were performed between the dependent variables of the size spectra slope with the independent variables of the annual gulf stream index, the corresponding regional commercial landings and sea surface temperature anomalies, and both small and large zooplankton indices. This was performed for annual lags up to 10 years. Significant lags identified in the CCF analysis were then included as additional candidate predictors in the multiple regression analysis of spectrum slope changes.

Non-Stationary Behavior Across Response Regimes

To highlight potential signatures of any non-stationary or non-linear discontinuous dynamics, a hallmark of regime-shift dynamics (Blocker 2023), we explored discontinuous changes in the size spectra using breakpoint analyses. Breakpoint analysis was performed on the annual time series of spectrum slopes to explore their temporal structures for evidence of any potential regime shifts. Breakpoint analyses were performed using the envcpt package (Killick et al. 2021). This package applies an automatic model selection process testing support for candidate model structures including constant/piecewise changes to the mean, variance, trends, & autocorrelations and any identified changepoints located using the pruned exact linear time algorithm (Rikardsen 2004). Best supported model structures from this procedure were then used to inform whether (if any) regional spectra expressed distinct periods of behavior. In the events where changepoints were identified, this was used as evidence that there may have been a community regime change. Following this thinking it is plausible for there to be inconsistencies or non-stationarity in the predictors’ impact on the community spectra, as different community regimes may be differently vulnerable to top-down pressures like commercial fishing. To allow for changes in these relationships across different regimes, an additional interaction term was added to the regression model candidates for that region.

Regime Specific Correlations

Based on any breakpoints detected, single driver correlations were prepared to visually explore how each driver was correlated within the different regimes. Additionally, spectra slope changes were also correlated with the body-size changes of each functional group as a window into what body-size changes might be informing the change in the size structure.

Multiple Regression Modeling of Hypothesized Drivers

Following these explorations into temporal structures, multiple regression models were developed and evaluated independently for each region. Potential drivers included in the models were the annual Gulf Stream Index, the annual regionally averaged SST anomaly value, regionally averaged small and large copepod indices, and the scaled annual commercial landings from that region. Gulf stream index, zooplankton indices, and SST anomalies were not standardized prior to use in model fitting and evaluation as the two former are already an index value and the latter is scaled to a thirty-year climatological reference period. In the event that CCF analysis identified an important time lag for a region’s independent variables, these specific time-lagged predictors were included as predictors in the candidate models, otherwise time-lags on predictors were not evaluated. For any region(s) that showed support for changepoints in their temporal structure, the periods separated by those breaks were included as an interaction term on all other predictors. In the absence of any identified changepoints, no multi-year periods were included as potential regime impacts. All candidate models for a region were evaluated using AIC and AIC weights to rank the most parsimonious models for each region’s dynamics.

Results

Abundance Changes

Stratified abundance estimates have been gradually increasing on the Northeast Shelf since 1970. Abundance growth accelerated after 2007 to peak levels in 2014. Abundance then began declining, which continued through 2019 (Figure 2).

In the Gulf of Maine, fish abundance remained relatively-low until the 1990’s, at which point it began to steadily rise–closely aligning with the pattern for the Northeast Shelf. This increase in abundance reversed briefly during 2002-2006, but continued to rise after 2006 before hitting a regional peak in 2016. Populations then declined through 2019 similar to the shelf-wide trend. Georges Bank abundances were consistently low from 1970 around 2010. By 2014 abundance had roughly quadrupled, propelled by strong recruitment classes of haddock. After 2014, abundance quickly fell to numbers more similar to the 1970-2010 levels by the end of the decade. Community abundance in Southern New England displayed higher inter-annual changes across all years compared to both Gulf of Maine and Georges Bank. Abundance in this area showed a less dramatic rise and fall than the Northern regions. Abundances began increasing rapidly here in 2007, before falling back to earlier levels by the end of the 2010’s decade. The Mid Atlantic Bight displayed the most inter-annual variability and had no major trends in fish abundance.

Regional Abundances

Groundfish species were the dominant functional group driving the abundance and biomass trends in the Gulf of Maine and Georges Bank, with the two southern regions showing a more balanced abundance distribution among the five functional groups (Figure 2).

Total Abundance by Individual Body-Size

Biomass Changes

Similar to abundance, the overall biomass was highest in the two northern regions, the Gulf of Maine and Georges Bank. Roughly half of the biomass sampled in these regions can be attributed to groundfish species, with the second largest contributions coming from elasmobranchs. Within the groundfish biomass, larger individuals >2kg in particular, declined during the 70’s and 80’s in these regions, never truly recovering. Beginning in the 2000’s there were signs that groundfish abundances were increasing as evidenced by increasing numbers of smaller individuals, however in both regions this trend appears to have reversed by the mid 2010’s. Elasmobranch biomass increased steadily throughout the survey time period across all regions, with the exception of southern New England. This area showed large 5-10 year swings in biomass, but no clear long-term trend. Larger elasmobranch were rare in all regions except for a period spanning the late 70’s through the early 90’s isolated to Georges Bank. Demersal species biomass was highest in the Gulf of Maine, dwarfing their contributions in other regions. Their biomass declined in the 70’s, was flat until the late 90’s, remaining relatively high until declining in the late 2010’s. Pelagic species biomass was low in all regions, and is unlikely to be representative of true biomass trends due to gear selectivity.

Regional Biomass

There was a distinct difference between Northern and Southern regions in the way biomass was distributed among the different functional groups. The largest contributors to biomass in the southern regions (southern New England & mid-Atlantic bight) was the elasmobranch community. While the northern regions (Gulf of Maine & Georges Bank) each had similar quantities of elasmobranch biomasses, there was also a comparable contribution of groundfish and in the Gulf of Maine there was a major component of demersal species as well.

Total Biomass by Individual Body-Size

Regional Size Spectra

At the start of our time series, back in the 1970’s, there was a clear difference in the relative positions of spectra parameters among the different regions. Gulf of Maine and Georges Bank showed the least steep spectra slopes in the earlier time periods with slopes around -1 & -1.1 respectively. The relatively flat slopes in these regions both steepened over time, settling near -1.3 (GoM) and -1.5 (GB). Gulf of Maine experienced much of its decline during the 1980’s and 1990’s. There was a brief reversal in this trend during the 2000’s, but slopes continued to steepen by 2010 and remained steep through 2019. Georges Bank did not experience as rapid of a decline, but experienced a similar long-term steepening. In contrast to the northern regions, SNE and MAB had steeper slopes in the -1.2 to -1.5 territory. The long term pattern for SNE was one of increasing volatility, but not so much a decline. The spectra slope for the MAB was less volatile, but similarly maintained a relatively stable wander around -1.4. By the end of the study period all regions had slopes that were at or near a similar level.

Size Spectra Drivers

Lagged Impacts of Predictors

Exploratory analysis on potentially important predictor time lags through the use of cross correlation functions raised 6 potential lagged-driver candidates for the driver regression analyses. CCF identified potential time lags in Georges Bank predictors, and in the Mid-Atlantic Bight. Lagged predictors added to Georges Bank regression model selection process included: A 2-year lag on the small zooplankton index, a 1-year lag on the large zooplankton index, a 1-year lag of SST, 4-year lag & 1-year lags of the GSI, and 4 & 5 year lags of commercial landings. Lagged predictors added to Mid-Atlantic Bight regression model selection process included: 5 & 7 year lags on the small zooplankton index and a 2-year lag on SST.

Spectra Slope Changepoints

Breakpoint analyses showed support for ecological regime change in one region, the Gulf of Maine. Evidence of breakpoints were found in 1999 & 2007 – when slope values increased briefly, before reversing course and falling further. The best-supported breakpoint model structure included breaks in the linear trend at 1999 & 2007, and a 2-year autocorrelation term. Suggesting that the community spectra behaved differently over 3 regimes with strong autocorrelation to its state two years prior. For Georges Bank spectra slopes showed a single trend structure, with slopes steepening over time. This was also the case for the Mid-Atlantic Bight, but with slopes becoming more shallow with time. In Southern New England a single mean (intercept) model best reflected the lack of change. With no strong support for either a long-term trend or break points. Based on these exploratory results, three regime periods were added to the Gulf of Maine’s dataset for multiple regression analysis. The three regimes were 1982-1998, 1999-2006, 2007-2019. These regimes were included as an interaction term with each of the original hypothesized drivers, permitting potentially different influences of the drivers within each regime. The two-year lag on the spectrum slope was also included based on its support in the most-parsimonious changepoint analysis model. No breakpoints or autocorrelation terms were added to models for the other regions.

Region Most Supported Model
gom trendar2cpt
gb trend
sne meancpt
mab trend

Change point analyses suggest there were as many as 3 distinct regimes in the Gulf of Maine slope changes. For all other regions there was not strong support for change points.

Spectra Slope Regression Analysis

WORKING HERE - Effect Reporting

fix the coefficients so the interactions now reflect the net result

Model rankings using AICc & delta-AICc (Figure 5.) for the Gulf of Maine’s size spectrum slope models best support a regression model containing two predictors with a stationary effect across all years, and two predictors whose effects across were allowed to vary across the regime periods identified by the changepoint analysis (non-stationary effects). The two stationary drivers impacts were the negative impact of the small zooplankton index (p = 0.011) and a negative impact from the two-year autocorrelation term highlighted by the exploratory changepoint analysis (p = 0.033). The inclusion of an interaction term between the three regimes (1982-1998, 1999-2006, & 2007-2019) and both the commercial landings index and the large zooplankton index allowed these drivers to differently impact size spectrum slope within the three regimes. Over those three periods the impact of landings during the first regime was positive (p = 0.009), the 1999-2006 regime was negative (p = 0.032), and during the third regime there was no relationship (p = 0.088). For the large zooplankton index during the first two regimes there was no significant impact (p = 0.2, p = >0.9), however during the third regime large zooplankton had a positive effect on the spectrum slope (p = <0.001). None of the terms for the regime periods themselves were significant in the final model. Models containing only stationary effects on the predictors were not retained by model selection. This agreed with visual assessments of their predictive performance and residual trends.

term Model ID 1982-1998 1999-2006 2007-2019 regime_2_net_change regime_3_net_change
(Intercept) Model 1 -1.2802667 -0.0431941 -1.8676802 -1.3234608 -3.1479470
landings Model 1 0.0550559 -0.1978798 -2.0444684 -0.1428239 -1.9894124
zp_large Model 1 -0.0647457 0.0082606 0.4033975 -0.0564851 0.3386518
(Intercept) Model 2 -0.9866921 -0.0684091 -0.4593722 -1.0551011 -1.4460643
landings Model 2 0.0628658 -0.2974706 -0.5871663 -0.2346048 -0.5243005
(Intercept) Model 3 -1.2371378 -0.0736346 -1.5117449 -1.3107723 -2.7488826
landings Model 3 0.0511718 -0.2180451 -1.6317484 -0.1668733 -1.5805766
zp_large Model 3 -0.0685200 0.0351875 0.4361328 -0.0333326 0.3676127
Characteristic Beta 95% CI1 p-value
blag2 -0.33 -0.64, -0.03 0.033
landings 0.06 0.01, 0.10 0.009
regime
    yrs_1982-1998
    yrs_1999_2006 -0.04 -0.15, 0.06 0.4
    yrs_2007_2019 -1.9 -4.1, 0.40 0.10
zp_large -0.06 -0.17, 0.04 0.2
zp_small -0.07 -0.12, -0.02 0.011
landings * regime
    landings * yrs_1999_2006 -0.20 -0.38, -0.02 0.032
    landings * yrs_2007_2019 -2.0 -4.4, 0.33 0.088
regime * zp_large
    yrs_1999_2006 * zp_large 0.01 -0.18, 0.19 >0.9
    yrs_2007_2019 * zp_large 0.40 0.18, 0.62 <0.001
0.653
Adjusted R² 0.515
Sigma 0.034
Statistic 4.71
p-value <0.001
df 10
Log-likelihood 77.3
AIC -131
BIC -112
Deviance 0.029
Residual df 25
No. Obs. 36
1 CI = Confidence Interval
Predictor GVIF Df GVIF^(1/(2*Df)) Interacts With Other Predictors
regime 1.000000e+00 17 1.000000 SST, GSI, landings, zp_small, zp_large --
SST 5.960859e+07 5 5.991435 regime GSI, landings, zp_small, zp_large
GSI 2.074176e+07 5 5.391189 regime SST, landings, zp_small, zp_large
landings 4.268892e+03 5 2.306915 regime SST, GSI, zp_small, zp_large
zp_small 1.460115e+08 5 6.552978 regime SST, GSI, landings, zp_large
zp_large 1.918367e+07 5 5.349253 regime SST, GSI, landings, zp_small

The best supported model for Georges Bank retained two predictors: sea surface temperature anomalies (p = 0.028) and the small zooplankton index at a 2-year lag (p = 0.006). With both the SST anomalies and the lagged small zooplankton index having negative effects on size spectrum slope. Five candidate models were within a delta-AIC range of <2. These top-models retained only bottom-up drivers as the best predictors, usually as some pair containing the small zooplankton at 0-2 year lags and either SST anomalies or the Gulf Stream Index. Suggesting that this region’s size spectrum changes are most highly correlated with environmental forces and not commercial landings.

Characteristic Beta 95% CI1 p-value
SST -0.03 -0.06, 0.00 0.028
zpslag2 -0.08 -0.13, -0.02 0.006
0.307
Adjusted R² 0.264
Sigma 0.052
Statistic 7.09
p-value 0.003
df 2
Log-likelihood 55.6
AIC -103
BIC -97.0
Deviance 0.085
Residual df 32
No. Obs. 35
1 CI = Confidence Interval
Predictor VIF
landings 1.929115
zpslag1 2.791261
SST 3.077146
zp_large 3.265561
zp_small 3.887965
zpslag2 4.634856
GSI 4.918679

Exploratory analysis for Southern New-England was suggestive of a lack of trend in size spectrum slopes. This was further confirmed by the model selection process. The best supported model of Southern New England retained only commercial landings, however this relationship was not significant (p = 0.13) and had very low performance (r-squared = 0.06).

Characteristic Beta 95% CI1 p-value
landings 0.01 0.00, 0.03 0.13
0.062
Adjusted R² 0.036
Sigma 0.046
Statistic 2.37
p-value 0.13
df 1
Log-likelihood 64.4
AIC -123
BIC -118
Deviance 0.075
Residual df 36
No. Obs. 38
1 CI = Confidence Interval
Predictor VIF
zp_large 1.507136
zp_small 1.621108
GSI 1.932934
landings 2.242318
SST 2.752092

Three top models were selected for the Mid-atlantic Bight region, all with non-stationary predictors. The best supported model showed that increases in the small zooplankton index had a negative impact on spectrum slope (p = 0.020). This result was present in the other top candidates, which each also included a negative correlation to large zooplankton or with commercial landings. Model performance was low among the top models (r-squared 0.14-0.16).

Characteristic Beta 95% CI1 p-value
zp_small -0.07 -0.13, -0.01 0.020
0.142
Adjusted R² 0.118
Sigma 0.060
Statistic 5.96
p-value 0.020
df 1
Log-likelihood 53.9
AIC -102
BIC -96.9
Deviance 0.130
Residual df 36
No. Obs. 38
1 CI = Confidence Interval
Predictor VIF
landings 1.331770
zp_large 1.373941
zp_small 1.636199
SST 1.763223
GSI 1.831599

Model Selection Results

Discussion

Potential Drivers Timeseries:

The following panels show the historical changes in each of the drivers. Landings have been scaled by average total landings within each region across all years. SST and GSI have not been scaled. This is different from how they are implemented in the regression analyses, when the landings were scaled over the 1982-2019 period.

Supplemental Materials

Table 1:

Common and scientific names for the species that constitute each functional group used in our analyses. X markers are used to indicate which regions each species has been caught in the data.

Functional Group Assignments and Regional Presence/Absence
Common Name Scientific Name Georges Bank Gulf of Maine Mid-Atlantic Bight Southern New England
Coastal - (18)
Atlantic Croaker micropogonias undulatus X X X
Atlantic Spadefish chaetodipterus faber X
Atlantic Thread Herring opisthonema oglinum X X
Black Sea Bass centropristis striata X X X X
Blackbelly Rosefish helicolenus dactylopterus X X X X
Blueback Herring alosa aestivalis X X X X
Bluefish pomatomus saltatrix X X X X
Butterfish peprilus triacanthus X X X X
Cunner tautogolabrus adspersus X X X X
Greater Amberjack seriola dumerili X X
Northern Kingfish menticirrhus saxatilis X X X
Scup stenotomus caprinus X X X X
Southern Kingfish menticirrhus americanus X
Spanish Mackerel scomberomorus maculatus X
Spanish Sardine sardinella aurita X
Spot leiostomus xanthurus X X
Striped Bass morone saxatilis X X X X
Weakfish cynoscion regalis X X X
Diadromous - (2)
American Shad alosa sapidissima X X X X
Atlantic Sturgeon acipenser oxyrhynchus X
Elasmobranch - (19)
Atlantic Angel Shark squatina dumeril X
Atlantic Sharpnose Shark rhizoprionodon terraenovae X
Barndoor Skate dipturus laevis X X X X
Bullnose Ray myliobatis freminvillei X X
Chain Dogfish scyliorhinus retifer X X X
Clearnose Skate raja eglanteria X X
Cownose Ray rhinoptera bonasus X
Little Skate leucoraja erinacea X X X X
Rosette Skate leucoraja garmani X X X X
Roughtail Stingray dasyatis centroura X
Sand Tiger carcharias taurus X
Sandbar Shark carcharhinus plumbeus X X
Smooth Butterfly Ray gymnura micrura X
Smooth Dogfish mustelus canis X X X X
Smooth Skate malacoraja senta X X X X
Spiny Butterfly Ray gymnura altavela X
Spiny Dogfish squalus acanthias X X X X
Thorny Skate amblyraja radiata X X X X
Winter Skate leucoraja ocellata X X X X
Groundfish - (25)
Acadian Redfish sebastes fasciatus X X X X
American Plaice hippoglossoides platessoides X X X X
Atlantic Cod gadus morhua X X X X
Atlantic Halibut hippoglossus hippoglossus X X X
Atlantic Wolffish anarhichas lupus X X X
Cusk brosme brosme X X X X
Fawn Cusk-Eel lepophidium profundorum X X X X
Fourspot Flounder paralichthys oblongus X X X X
Goosefish lophius americanus X X X X
Haddock melanogrammus aeglefinus X X X X
Longhorn Sculpin myoxocephalus octodecemspinosus X X X X
Northern Searobin prionotus carolinus X X X X
Ocean Pout macrozoarces americanus X X X X
Offshore Hake merluccius albidus X X X X
Pollock pollachius virens X X X X
Red Hake urophycis chuss X X X X
Sea Raven hemitripterus americanus X X X X
Silver Hake merluccius bilinearis X X X X
Spotted Hake urophycis regia X X X X
Summer Flounder paralichthys dentatus X X X X
White Hake urophycis tenuis X X X X
Windowpane Flounder scophthalmus aquosus X X X X
Winter Flounder pseudopleuronectes americanus X X X X
Witch Flounder glyptocephalus cynoglossus X X X X
Yellowtail Flounder limanda ferruginea X X X X
Pelagic - (4)
Atlantic Herring clupea harengus X X X X
Atlantic Mackerel scomber scombrus X X X X
Buckler Dory zenopsis conchifera X X X X
Round Herring etrumeus teres X X X X
Functional group assignments adapted from Hare et al. 2010
Top Commercial Fisheries Landings of Northeastern US (by weight)
Avg. Annual Landings (lb.) Total Landings (lb.) Total Value ($)
Gulf of Maine - 1960
Hake, Silver 16.58M 281.87M 8.71M
Herring, Atlantic 11.57M 138.83M 2.50M
Redfish, Acadian 2.12M 88.97M 3.41M
Gulf of Maine - 1970
Herring, Atlantic 22.78M 501.08M 19.70M
Menhaden, Atlantic 17.78M 373.48M 7.87M
Redfish, Acadian 3.14M 219.85M 23.87M
Gulf of Maine - 1980
Herring, Atlantic 21.78M 653.26M 34.52M
Menhaden, Atlantic 21.24M 509.75M 12.77M
Pollock 3.33M 229.57M 62.00M
Gulf of Maine - 1990
Herring, Atlantic 25.21M 958.12M 54.12M
Cod, Atlantic 2.35M 138.76M 131.76M
Shark, Dogfish, Spiny 3.34M 120.17M 15.95M
Gulf of Maine - 2000
Herring, Atlantic 2.99M 47.77M 4.31M
Monkfish/Angler/Goosefish 716.21K 31.51M 51.13M
Cod, Atlantic 692.95K 30.49M 42.30M
Gulf of Maine - 2010
Tuna, Bluefin 209.06K 3.76M 33.30M
Shark, Dogfish, Spiny 479.11K 2.87M 590.62K
Pollock 188.20K 1.69M 2.08M
Georges Bank - 1960
Haddock 15.00M 270.06M 34.41M
Hake, Silver 6.83M 95.57M 3.19M
Cod, Atlantic 4.88M 87.89M 8.12M
Georges Bank - 1970
Cod, Atlantic 7.78M 233.48M 59.16M
Flounder, Yellowtail 4.62M 138.52M 43.16M
Redfish, Acadian 2.63M 76.37M 9.09M
Georges Bank - 1980
Cod, Atlantic 10.11M 404.40M 211.60M
Flounder, Winter 2.50M 100.11M 89.84M
Haddock 2.36M 94.27M 66.68M
Georges Bank - 1990
Cod, Atlantic 4.27M 192.29M 190.26M
Hake, Silver 1.79M 76.82M 20.49M
Flounder, Winter 1.23M 56.43M 75.59M
Georges Bank - 2000
Cod, Atlantic 2.17M 62.91M 75.20M
Herring, Atlantic 3.49M 48.92M 3.73M
Haddock 1.54M 43.01M 55.37M
Georges Bank - 2010
Hake, Silver 155.88K 779.40K 499.90K
Haddock 39.65K 118.95K 143.04K
Flounder, Winter 40.40K 80.80K 216.28K
Southern New England - 1960
Other Fish, Bony 14.84M 400.77M 3.73M
Flounder, Yellowtail 6.56M 196.83M 19.12M
Flounder, Winter 2.52M 70.58M 7.01M
Southern New England - 1970
Menhaden, Atlantic 9.99M 239.84M 5.12M
Other Fish, Bony 4.05M 206.59M 2.49M
Flounder, Yellowtail 2.07M 153.55M 36.47M
Southern New England - 1980
Menhaden, Atlantic 6.60M 217.68M 10.21M
Hake, Silver 2.56M 205.02M 46.11M
Flounder, Yellowtail 1.66M 132.92M 83.38M
Southern New England - 1990
Hake, Silver 2.52M 196.81M 78.54M
Herring, Atlantic 2.12M 129.02M 7.19M
Menhaden, Atlantic 3.71M 125.98M 8.69M
Southern New England - 2000
Mackerel, Atlantic 2.55M 135.06M 15.60M
Hake, Silver 1.00M 55.25M 26.89M
Skate, Nk 950.56K 49.43M 6.53M
Southern New England - 2010
Scup 161.10K 6.44M 4.29M
Hake, Silver 145.07K 4.21M 3.12M
Flounder, Summer 80.43K 3.86M 11.54M
Mid-Atlantic Bight - 1960
Flounder, Summer 2.03K 4.05K 720.00
Flounder, Yellowtail 2.33K 2.33K 214.00
Flounder, Witch 395.00 395.00 36.00
Mid-Atlantic Bight - 1970
Menhaden, Atlantic 10.20M 50.98M 1.59M
Weakfish/Sea Trout, Squeteague 886.91K 9.76M 1.40M
Scup 876.60K 8.77M 2.09M
Mid-Atlantic Bight - 1980
Menhaden, Atlantic 30.78M 646.41M 10.94M
Flounder, Summer 1.15M 83.83M 72.00M
Scup 550.89K 37.46M 15.53M
Mid-Atlantic Bight - 1990
Menhaden, Atlantic 115.86M 4.63B 286.14M
Mackerel, Atlantic 1.67M 103.62M 13.87M
Croaker, Atlantic 1.35M 71.65M 22.53M
Mid-Atlantic Bight - 2000
Menhaden, Atlantic 69.60M 2.64B 167.17M
Croaker, Atlantic 2.16M 106.02M 42.93M
Mackerel, Atlantic 1.70M 59.41M 6.38M
Mid-Atlantic Bight - 2010
Menhaden, Atlantic 118.29M 1.89B 154.46M
Bass, Striped 1.70M 25.56M 75.05M
Croaker, Atlantic 1.08M 24.81M 21.37M
Landings data obtained from the Greater Atlantic Regional Fishing Office (GARFO)

Single Driver Correlations

References

Andersen, K., and Beyer, J. E. 2006. Asymptotic size determines species abundance in the marine size spectrum. The American Naturalist, 168: 54–61.
Atkinson, D. 1995. Effects of temperature on the size of aquatic ectotherms: Exceptions to the general rule. Journal of Thermal Biology, 20: 61–74. https://linkinghub.elsevier.com/retrieve/pii/030645659400028H.
Blackburn, T. M., Gaston, K. J., and Loder, N. 1999. Geographic gradients in body size: a clarification of Bergmann’s rule. BIODIVERSITY RESEARCH. Diversity <html_ent glyph="@amp;" ascii="&amp;"/> Distributions, 5: 165–174. http://dx.doi.org/10.1046/j.1472-4642.1999.00046.x.
Brown, J., West, G. B., and Enquist, B. 2000. Scaling in biology: Patterns and processes, causes and consequences. In pp. 1–24.
Findlay, H. S., and Turley, C. 2021. Chapter 13 - Ocean acidification and climate change. In pp. 251–279. Ed. by T. M. Letcher. Elsevier. https://www.sciencedirect.com/science/article/pii/B978012821575300013X.
Frederikse, T., Landerer, F., Caron, L., Adhikari, S., Parkes, D., Humphrey, V. W., Dangendorf, S., et al. 2020. The causes of sea-level rise since 1900. Nature, 584: 393–397. http://dx.doi.org/10.1038/s41586-020-2591-3.
Glantz, M. H., Katz, R. W., and Nicholls, N. (Eds). 1991. Teleconnections linking worldwide climate anomalies: scientific basis and societal impact. Cambridge University Press, Cambridge [England] ; New York.
Hillaert, J., Hovestadt, T., Vandegehuchte, M. L., and Bonte, D. 2018. Size-dependent movement explains why bigger is better in fragmented landscapes. Ecology and Evolution, 8: 10754–10767. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6262741/.
Kerr, S. R., and Dickie, L. M. 2001. The Biomass Spectrum: A Predator-Prey Theory of Aquatic Production. Columbia University Press.
Liu, Z., and Alexander, M. 2007. Atmospheric bridge, oceanic tunnel, and global climatic teleconnections. Reviews of Geophysics, 45. https://onlinelibrary.wiley.com/doi/abs/10.1029/2005RG000172.
Meyssignac, B., Boyer, T., Zhao, Z., Hakuba, M. Z., Landerer, F. W., Stammer, D., Köhl, A., et al. 2019. Measuring global ocean heat content to estimate the earth energy imbalance. Frontiers in Marine Science, 6. http://dx.doi.org/10.3389/fmars.2019.00432.
Neto, A. G., Langan, J. A., and Palter, J. B. 2021. Changes in the Gulf Stream preceded rapid warming of the Northwest Atlantic Shelf. Communications Earth & Environment, 2: 1–10. https://www.nature.com/articles/s43247-021-00143-5.
Nye, J., Link, J., Hare, J., and Overholtz, W. 2009. Changing spatial distribution of fish stocks in relation to climate and population size on the Northeast United States continental shelf. Marine Ecology Progress Series, 393: 111–129. http://www.int-res.com/abstracts/meps/v393/p111-129/.
Pershing, A. J., Alexander, M. A., Hernandez, C. M., Kerr, L. A., Le Bris, A., Mills, K. E., Nye, J. A., et al. 2015. Slow adaptation in the face of rapid warming leads to collapse of the gulf of maine cod fishery. Science, 350: 809–812. https://www.science.org/doi/full/10.1126/science.aac9819.
Pinsky, M. L., Worm, B., Fogarty, M. J., Sarmiento, J. L., and Levin, S. A. 2013. Marine Taxa Track Local Climate Velocities. Science, 341: 1239–1242. https://science.sciencemag.org/content/341/6151/1239.
Sheldon, R. W., Prakash, A., and Sutcliffe Jr., W. H. 1972. The Size Distribution of Particles in the Ocean1. Limnology and Oceanography, 17: 327–340. https://onlinelibrary.wiley.com/doi/abs/10.4319/lo.1972.17.3.0327.
Shin, Y.-J., Rochet, M.-J., Jennings, S., Field, J. G., and Gislason, H. 2005. Using size-based indicators to evaluate the ecosystem effects of fishing. ICES Journal of Marine Science, 62: 384–396. https://doi.org/10.1016/j.icesjms.2005.01.004.
Sprules, W. G., and Barth, L. E. 2016. Surfing the biomass size spectrum: some remarks on history, theory, and application. Canadian Journal of Fisheries and Aquatic Sciences, 73: 477–495. http://www.nrcresearchpress.com/doi/10.1139/cjfas-2015-0115.
University of South Carolina, Meyer-Gutbrod, E., Greene, C., Davies, K., and Johns, D. 2021. Ocean Regime Shift is Driving Collapse of the North Atlantic Right Whale Population. Oceanography, 34: 22–31. https://tos.org/oceanography/article/ocean-regime-shift-is-driving-collapse-of-the-north-atlantic-right-whale-population.
Warming trends increasingly dominate global ocean | nature climate change. (n.d.). https://www.nature.com/articles/s41558-020-0822-0.
White, E. P., Ernest, S. K. M., Kerkhoff, A. J., and Enquist, B. J. 2007. Relationships between body size and abundance in ecology. Trends in Ecology & Evolution, 22: 323–330. https://linkinghub.elsevier.com/retrieve/pii/S0169534707000985.
White, E. P., Enquist, B. J., and Green, J. L. 2008. On Estimating the Exponent of Power-Law Frequency Distributions. Ecology, 89: 905–912. https://onlinelibrary.wiley.com/doi/abs/10.1890/07-1288.1.